Playing with the Agent-Agent network for the ABM

Initially, the model considers a complete Agent-Agent interaction network.

We start with two identical models initialized with the exact same seed so that

• Agents are in the same positions
• Media are in the same positions
• Influencers are in the same positions

All other parameters being exactly equal.

Everyone interacts with everyone

Everyone interacts only within their clicque

It’s hard to tell a difference, they both look almost the same. The one with the echo chamber effect exhibits a “lateral spreading”.

The idea was to use Wasserstein distances to measure how different the distributions are.

• Last frame of normal simulation
• Last frame of echo chamber simulation

Conclusion (?)

• The main effect is still the attraction excerted by media outlets

• The spreading may have to do with how the agents feel attraction to their influencer.

• The spreading differences persist even if I move the media agencies closer to the center.